Absence of regional brain volume change in schizophrenia associated with short-term atypical antipsychotic treatment

1. Background 

Longitudinal MRI studies in schizophrenic patients have been performed primarily after periods of antipsychotic treatment greater than twelve weeks. There is a paucity of information about regional brain volume change associated with periods of antipsychotic treatment less than twelve weeks. Brain volume changes that occur over brief time periods are likely to be potentially reversible and are unlikely to reflect fixed structural abnormalities (McClure et al., 2006). A global objective of this pilot study was to reduce the paucity of scientific inquiry into the association of short-term antipsychotic treatment with regional brain volume change in schizophrenia.

Increased caudate volume in antipsychotic-naïve schizophrenia patients has been associated with typical antipsychotic treatment for time periods greater than twelve weeks (Chakos et al., 1995, Chakos et al., 1994, Corson et al., 1999a, Corson et al., 1999b, Dazzan et al., 2004, Dazzan et al., 2005, DeLisi et al., 1994, Delisi et al., 1991, DeLisi et al., 1997, Gur et al., 1998, Heitmiller et al., 2004, Kubicki et al., 2002, Keshavan et al., 1998, Keshavan et al., 1994, Lang et al., 2001, Mathalon et al., 2001, Pressler et al., 2005, Scheepers et al., 2001, Shihabuddin et al., 1998, Tamagaki et al., 2005, Tauscher-Wisniewski et al., 2005). Increased caudate volume in rats treated with atypical antipsychotics (Andersson et al., 2002) has also been reported. Few investigators (Christensen et al., 2004, Garver et al., 2005, Massana et al., 2005, McClure et al., 2006, Strungas et al., 2003, Tauscher-Wisniewski et al., 2005) have examined volume change in the caudate following briefer antipsychotic treatment periods (Massana et al., 2005, Tauscher-Wisniewski et al., 2005). We sought to determine if regions of longitudinal volume change could be identified in the caudate following periods of antipsychotic treatment of twelve weeks or less in medication-free chronic schizophrenia patients. Specifically, we measured change in caudate volume in a small sample of chronic schizophrenic inpatients treated with atypical antipsychotics for twelve weeks following medication washout for a mean of 39.4 days.

In the past, manual region-of-interest analysis (ROI) was considered the “gold standard” for determining regional volume with structural MRI studies (McCarley et al., 1999). The limitations of manual ROI analysis include the following. First, volume can be determined only for the ROIs drawn, providing no information on the regions that are not examined. Second, variability due to inter-rater reliability can be high. Third, elimination of inter-rater reliability requires extensive, time-consuming, and expensive rater training, which is beyond the capacity of many individual investigators. Voxel-based morphometry (VBM), a newer method for the analysis of structural magnetic resonance images, is not hampered by these limitations, because it does not introduce inter-rater bias and the entire brain is sampled in an unbiased manner. Since VBM requires estimation of smooth, low frequency deformation fields, it is within budgetary capabilities of many research units. There are some limitations to VBM (Mechelli et al., 2005). First, shape differences attributable to misregistration during spatial normalization, rather than actual group differences, can be detected. Second, global rather than local regional volume change may be more accurately detected, because of the relatively imprecise registration used. Third, the accuracy of localization is negatively affected by smoothing, which may shift the peak of the SPM towards regions of low variance. Fourth, spatially complex, subtle, or changes that are related to changes that occur elsewhere in the brain may not be detected by this mass univariate approach. Fifth, the exact nature of tissue changes identified with VBM is still poorly understood.

Consequently, the results of longitudinal MRI studies using VBM vary considerably(Ananth et al., 2002, Hulshoff et al., 2001, Job et al., 2002, Kubicki et al., 2002, Pantelis et al., 2002, Sowell et al., 2000, Thompson et al., 2001, Wilke et al., 2001, Wright et al., 1999a, Wright et al., 1999b) and their interpretation is difficult. VBM as it is performed in theses studies and this study detects the presence and location of volume change, but not the magnitude. Validation of voxel-based studies has not been extensive, limited to a few brain regions, experimental conditions, and analysis methods. Recently, newer semi- and fully-automated methods of analyzing regional brain volume have been developed, since the advent of VBM. “Snake evolution” is one of these recently developed methodologies. “Snake” refers to a closed curve or surface in three dimensions that represents a segmentation. The snake begins as a very rough estimate of the anatomical structure of interest and evolves to a very close approximation of the structure. The evolution of the snake is governed by a mathematical equation describing the velocity of every point on the snake at any particular time. The velocity of each point depends on the shape of the snake and on the intensities of the image in the neighborhood of the point. Snake evolution has been implemented in a freely available program known as ITK-SNAP (http://www.ia.unc.edu/dev/download/index.htm). Validation studies suggest that this method has excellent efficiency and reliability(Yushkevich et al., 2006), but there are no published studies comparing the results of ITK-SNAP to VBM. A sub-aim of this pilot study was to determine if these two methods of detecting volume change-ITK-SNAP and VBM-produced similar results in the caudate.

ITK-SNAP-VBM comparison was purposefully limited to the caudate for several reasons. First, reducing the chance of false positive findings from the examination of multiple regions of interest is particularly important in structural neuroimaging studies. In this pilot study, the likelihood of false positives was greatly reduced, since repeated-measures ANOVA performing was performed only on the left and right caudate. Second, maximizing the chance of obtaining a true positive finding is also particularly important in a structural neuroimaging study with a small sample size. The need for correction for multiple comparisons was avoided by limiting the manual ITK-SNAP analysis to left and right caudate only. Third, the caudate is an ideal candidate region to compare the results of image analysis methods. The direction of volume changes associated with atypical antipsychotic treatment can be directly predicted for the caudate, since the literature is more consistent then for other brain regions.

Cortical grey matter, white matter, and cerebrospinal fluid volume change in schizophrenic patients has been associated with typical antipsychotic treatment for periods greater than twelve weeks using both typical (DeLisi et al., 1988, DeLisi et al., 1992, DeLisi et al., 1995; Gur, 1998; Ho et al., 2003; Mathalon, 2001; McCormick et al., 2005) and atypical (Gogtay et al., 2004, McCormick et al., 2005, Molina et al., 2005) antipsychotics. The complete absence of statistically significant volume changes in these regions has been reported. The association of volume change in cortical grey matter, white matter, and cerebrospinal fluid and antipsychotic treatment is not consistent (Chakos et al., 1994, James et al., 2004, Keshavan et al., 1998, Pressler et al., 2005). The direction of reported volume changes are also not consistent (Lieberman et al., 2005, Nopoulos et al., 2005), even in studies adequately powered to detect volume change (Gogtay et al., 2004, Ho et al., 2003, McCormick et al., 2005, Molina et al., 2005, Mathalon et al., 2001, Gur et al., 1998, DeLisi et al., 1994, Delisi et al., 1991, DeLisi et al., 1997, Gur et al., 1998, Lieberman et al., 2005).

It is also unclear if volume change in cortical grey matter, white matter or cerebrospinal fluid follows briefer periods of antipsychotic treatment, because so few investigators(Christensen et al., 2004, Garver et al., 2005, Massana et al., 2005, McClure et al., 2006, Strungas et al., 2003, Tauscher-Wisniewski et al., 2005; McClure, 2006; Strungas et al., 2003) have examined this question. A previous study from our research group demonstrated localized grey matter volume change in the frontal and temporal lobes following both withdrawal from, and chronic treatment with, antipsychotic medications (p<.05, uncorrected for multiple comparisons). The changes did not persist after adjustment for multiple corrections with the false discovery rate (McClure, 2006). The results of this study also demonstrated good correlation between the results of Bonferroni corrected manual ROI and false discovery rate (FDR) adjusted VBM methods in frontal and temporal grey matter. A secondary aim of our current study was to determine if regions of longitudinal volume change could be identified in grey matter, white matter, ventricular-sulcal regions following short periods of antipsychotic treatment in chronic schizophrenia patients. We accomplished this aim by identifying significant regions of tissue volume change measured with VBM in the aforementioned sample of chronic schizophrenic inpatients treated with atypical antipsychotics for twelve weeks following medication washout for a mean of 39.4 days.

Directional a priori hypothesize for antipsychotic-associated volume changes could be directly made for the caudate, based on previous findings described in the literature. Since reversal of caudate enlargement in schizophrenic patients previously treated with typical antipsychotics has been reported following treatment with atypicals (Chakos, 1995; Corson et al., 1999b, Dazzan et al., 2004; Gur, 1998; Scheepers, 2001) and no caudate volume change has also been reported after more than twelve weeks of atypical antipsychotic treatment (Heitmiller et al., 2004, Lang et al., 2001, Lang et al., 2004), we predicted that left and right caudate volume would either decrease or not change. We also predicted that either no change or decreases in grey matter volume would be observed in the caudate with VBM after adjustment for the false discovery rate. Based on the VBM results from our previous study (McClure, 2006), we predicted that if any localized frontal or temporal grey matter volume change were seen, it would not persist after false discovery rate adjustment. We hypothesized that no white matter and cerebrospinal fluid volume change would be observed after FDR adjustment. Although we did not expect changes in the larger grey, white and sulcal–cerebrospinal fluid regions, we expected VBM to sensitively detect these sorts of changes if they were present, since the method is designed to detect global volume change. With respect to the methods comparison, we predicted that localized caudate volume changes would be detected with the ITK-SNAP, but VBM methods would be less likely to detect localized caudate volume change.

2. Materials and methods 

2.1. Subjects 

Ten psychiatric inpatients at NIMH Clinical Center, Bethesda, Maryland enrolled voluntarily in a research protocol between 1998–2002. Subjects were independently determined to have capacity to give, and subsequently gave informed consent to participate in the research protocol, which followed current ethical standards and was approved by the NIMH Internal Review Board. Patient inclusion criteria were: (1.) normal brain MRI, defined as the absence of structural abnormality, prominent anatomical variants, or technical artifacts, including any motion detectable by human observation; (2.) absence of criteria for substance abuse six months before, or criteria for substance dependence one year before enrollment; (3.) absence of a seizure disorder or history of head trauma involving loss of consciousness; (4.) absence of any medical condition known to affect MRI of the brain; (5.) presence of DSM-IV criteria for schizophrenia or schizoaffective disorder(First et al., 1994). Diagnosis was confirmed by a board-certified psychiatrist following structured clinical interview for DSM-IV disorders. Patient exclusionary criteria included (1.) presence of another neurological condition (besides schizophrenia or schizoaffective disorder) or a medical condition that affects brain MRI.

As part of a larger research protocol designed to assess the effects of antipsychotic medications on neurocognitive and neuroimaging phenotypes, patients elected to withdraw from medications during a three to six week period of close psychiatric, nursing, and medical monitoring. After the period of antipsychotic withdrawal, subjects were restarted on antipsychotics. Two MRI studies were performed, at baseline off medications, and twelve weeks after restarting medications. Twelve subjects originally completed the protocol. Since there were only two subjects treated with typical antipsychotics, they were dropped from the analysis. Ten subjects, nine males and one female, were treated with second-generation atypical antipsychotics at clinical doses (olanzapine 10–20 mg per day, risperidone 4–6 mg per day, quetiapine 300–800 mg per day, clozapine 500–1000 mg per day). Other adjunctive treatments included cogentin 2 mg per day and valproic acid 1500 mg per day, and fluoxetine 40 mg per day in three different subjects respectively. All patients except one were previously treated with the same second-generation atypical antipsychotics. Ethnicity of the subjects was Caucasian of European decent. Of note, all ten subjects also participated in the previously mentioned study (McClure, 2006) and analysis of this follow up data was planned in an a priori manner. Time, resources and ethical considerations precluded recruitment of additional subjects. The demographic composition of the sample is described in Table 1. Premorbid IQ was estimated from the by Wide Range Achievement Test-Reading subscale (Wilkinson, 1993) administered by a neuropsychologist. Handedness was determined by the Edinburgh Handedness Inventory (Oldfield, 1971) by a trained research assistant.

Table 1. Demographic and clinical variables

		Atypicals			t-value	df	p
		Mean	Range	S.D.
	Sample size	N=10
	Gender	Male=9
		Female=1
	Age in years	36.7	25–46	7.7	1.8	7	0.12
	Duration of illness in years	10.6	1–26	9.7	−0.2	7	0.85
	Education (years)	14.4	12–20	3.3	−0.2	7	0.84
	Intelligence Quotient	92.7	78–102	9.1	−1.7	5	0.16
	Handedness Quotient	0.6	−0.8–1.0	0.69	2.2	6	0.07
	Days off medications	39.4	12–105	33	0.6	7	0.56

3. Results 

3.1. Caudate ITK-SNAP 

Mean right caudate volume off medications was 3.788 cc's (S.D.=0.598 cc's) and on medications was 3.724 cc's (S.D.=0.630 cc's). Change in mean right caudate volume did not approach statistical significance (p=.80, df,=1, F(1,7)=.0636). Mean left caudate off medication was 3.885 cc's (S.D.=0.509 cc's) and on medications was 3.608 cc's (S.D.=0.529 cc's). Similarly, change in left mean caudate volume also did not reach statistical significance (p=.253, df=1 F(1,7)=.254) (see Fig. 1, Fig. 2).

View Large Image Download to PowerPoint

Fig. 1. ANOVA right caudate.

View Large Image Download to PowerPoint

Fig. 2. ANOVA left caudate.

4. Discussion 

No significant caudate volume change was observed in medication-free chronic schizophrenic patients treated briefly with atypical antipsychotics. This finding stands in contrast to the findings of previous cross-sectional (Corson et al., 1999a, Corson et al., 1999b, Dazzan et al., 2005, Dazzan et al., 2004, Shihabuddin et al., 1998, Tauscher-Wisniewski et al., 2005) and long-term(Chakos, 1995; Corson et al., 1999a, Corson et al., 1999b, Keshavan et al., 1998, Lang et al., 2001, Lang et al., 2004, Massana et al., 2005, Scheepers et al., 2001) as well as short-term (Massana et al., 2005, Tauscher-Wisniewski et al., 2005, McClure et al., 2006, Christensen et al., 2004, Garver et al., 2005, Massana et al., 2005, McClure et al., 2006, Strungas et al., 2003, Tauscher-Wisniewski et al., 2005) longitudinal studies. Caudate volume change was most likely absent because the subjects were treated previously with atypical antipsychotics only, and because the treatment period was only a few months. It is possible that caudate volume change occurred in our subjects, but small sample size/low power precluded detection of very small volume changes. It is also possible that the results were influenced by confounding variables. There are many factors associated with physiological alterations in the brain, and hence brain volume change (Weinberger and McClure, 2002). For example, two of our subjects were treated with either a mood stabilizer (valproic acid) and an SSRI (fluoxetine), psychotropic medication that have been associated with brain volume change.

In marked contrast to the cortical grey matter volume changes reported in long-term (Gogtay et al., 2004, Gur et al., 1998, Lieberman et al., 2005; Mathalon, 2001; Molina et al., 2005, Nopoulos et al., 2005, Gur et al., 1998, DeLisi et al., 1994, Delisi et al., 1991, DeLisi et al., 1997, Gur et al., 1998) and short-term (Garver et al., 2005, Strungas et al., 2003, Christensen et al., 2004, Massana et al., 2005, McClure et al., 2006, Tauscher-Wisniewski et al., 2005) longitudinal studies, no volume change was seen in frontal or temporal grey matter regions where treatment effects were observed in our previous VBM study(McClure, 2006). Although volume changes in white matter have been previously observed in both long-term (Ho et al., 2003, Molina et al., 2005, Nopoulos et al., 2005, Gur et al., 1998, DeLisi et al., 1994, Delisi et al., 1991, DeLisi et al., 1997, Mathalon et al., 2001) and short-term (Christensen et al., 2004, Garver et al., 2005, Massana et al., 2005, McClure et al., 2006, Strungas et al., 2003, Tauscher-Wisniewski et al., 2005) longitudinal treatment studies, we did not observe an effect of treatment status in white matter. Our negative findings in third, fourth or lateral ventricle and cerebral cortical sulci, contrast with the findings of previous long-term (DeLisi et al., 1988, DeLisi et al., 1992, DeLisi et al., 1995, DeLisi et al., 1997; Mathalon, 2001; Gur et al., 1998, DeLisi et al., 1994, Delisi et al., 1991, DeLisi et al., 1997, 1998a,b; Mathalon et al., 2001, Gogtay et al., 2004, Ho et al., 2003, Lieberman et al., 2005, McCormick et al., 2005, Molina et al., 2005) and short-term (Christensen et al., 2004, Garver et al., 2005, Massana et al., 2005, McClure et al., 2006, Strungas et al., 2003, Tauscher-Wisniewski et al., 2005) longitudinal treatment studies examining volume change in cerebrospinal fluid containing structures.

The inconsistencies in both direction and region of findings of our study compared to previous studies are less perplexing when one considers that previous longitudinal studies have a unique pattern of findings (Marenco et al., 2002, McClure et al., 2006, Styner et al., 2005, Weinberger and McClure, 2002). Certainly, our findings are no more or less inconsistent than those of other investigators. Our findings are of considerable clinical significance given the proposed use of brain volume change as biomarkers or intermediate phenotypes. Our negative findings should be interpreted in light of the weaknesses of this pilot study, particularly the small sample size, and low power to detect small changes. The absence of random assignment of subjects is a potential weakness, although our subjects appear to have similar characteristics to other chronic schizophrenic patients. The absence of an appropriate control group could be considered a weakness. However, in designing the study, we considered what an appropriate control group would represent: a longitudinal sample of age-matched normal controls was deemed not to be appropriate, since they would not have not been treated with antipsychotics; obtaining a group of healthy controls treated with atypical antipsychotics, followed by a medication-free period of 4 weeks, and retreated with antipsychotics for three months, was neither possible nor ethical; data from the comparison group of chronically treated patients from our previous study (McClure, 2006) has already been published, so we did not consider it appropriate to publish the same data twice, and so; the within-subject repeated-measures design seemed to best address the experimental question. We cannot comment on the potential effects of gender on volume change related to antipsychotic treatment, which could be considered a weakness. We did not stratify our analysis by gender, since our sample was very small, and it did not seem appropriate to compare our sample to larger samples in other studies.

This report is the first comparison of ITK-SNAP to VBM methods in the caudate using the same MRI data. We observed little inconsistency between the two methods in the caudate. This is surprising, given that VBM and ITK-SNAP methods differ in their potential sources of variance, error, and sensitivity to volume change. It has been suggested that voxel-based morphometry is especially useful for detecting grey matter reductions that are very small, are located in areas with high variability of volume or are inconsistent in location (Wright et al., 1999a, Wright et al., 1999b), although this is not entirely consistent with its original design (Mechelli et al., 2005). Although VBM may be vulnerable to partial volume effects when grey or white matter voxels are misclassified (Job et al., 2002), and the location of maximal change may be altered by the use of smoothing (Kubicki et al., 2002), this method produced identical results when compared to a new, efficient, reliable user-operated tool.

Our results do not directly show that brief periods of atypical antipsychotic treatment are associated with basal ganglia and cortical volume change, but there is much evidence to suggest that such an association does exist. However, even with adequately powered unbiased studies, the presence and the direction of changes observed after short-term treatment are unlikely to be consistent in region or direction, as with studies of long-term treatment. Volume change during short-term antipsychotic treatment is likely to reflect transient physiological changes, such as those attributable to alteration in blood flow or metabolic demands. Our study makes a valuable contribution to the scientific literature because it is one of the few published that examines short-term atypical antipsychotic treatment and directly compares results of ITK-SNAP to voxel-based methods.

Role of funding source 

The funding for this study was provided by the Clinical Brain Disorders Branch, National Institutes of Mental Health, which provided for the study design and data collection. CBDB-NIMH had no role in the analysis and interpretation of data; in the writing of the report; and in the decision to submit the paper for publication. The STRT grant from NIDDK provided stipends for Stacy Greeter and Khary Carew, who performed data analysis, but NIDDK had no other role in the study.

Contributors 

Robert K. McClure designed the study with Dr. Weinberger, gathered the MRI data, performed the ROI and VBM analysis, statistical analysis, created the tables and figures, and wrote the manuscript.

Khary Carew and Stacy Greeter contributed equally and performed the caudate ROI analysis.

Emily Maushauer established methods for obtaining inter-rater reliability for the caudate ROIs.

Grant Steen assisted with preparation of the manuscript.

Daniel R. Weinberger designed the study with Dr. McClure.

Conflict of interest 

Dr. McClure has contracted with X Marketing Consulting LLC to do several non-promotional presentations for patients, families, advocates, and mental health professionals on behalf of Jannsen pharmaceuticals within the past three years.